Research Engineer — Agentic Audio ML Workflows

Bose Corporation•Foxborough, MA

4d•$165,400 - $227,500

About The Position

Bose’s team develops novel AI-powered audio processing algorithms designed to run in real time on physical devices. The Research Engineer will build agentic tools and automation to make audio machine learning research faster and more reproducible. The work focuses on orchestrating model training, evaluation, experimentation, and reporting across different AI models, using on-premises and/or cloud GPU infrastructure. The role involves close collaboration with audio researchers to convert repetitive experiments into reusable workflows and clear dashboards or reports.

Requirements

Degree in computer science, machine learning, electrical engineering, or a related field, or equivalent practical experience.
Strong software engineering and Python skills, with hands-on experience building LLM agents or agentic workflows that invoke tools and execute multi-step tasks.
Comfort tackling complex, open-ended problems that require experimenting with and discarding ideas before finding an effective solution.
Experience with ML training and evaluation pipelines, PyTorch, experiment tracking, and automated testing.
Practical experience operating on-premises Linux GPU machines, including CUDA environments, containers, job scheduling, and resource monitoring.
Working knowledge of audio ML and DSP, including common datasets, audio metrics, and model development workflows.
Ability to build reliable, observable systems and communicate effectively with research scientists.

Nice To Haves

Experience with orchestration frameworks, distributed training, CI/CD, or MLOps.

Responsibilities

Design and implement agentic workflows that plan, launch, monitor, diagnose, and summarize audio ML experiments, with appropriate human oversight.
Automate data preparation, model training, hyperparameter searches, benchmarking, regression testing, and result comparison across research projects.
Build reliable tooling for on-premises GPU scheduling and utilization, experiment tracking, artifact management, and reproducible execution.
Connect LLM-powered agents to scripts, repositories, datasets, evaluation suites, and compute resources through well-defined tools and safeguards.
Partner with audio researchers to turn repetitive experiments into reusable workflows and clear dashboards or reports.